EDA

• Train Dataset: 12,408건
• Test Dataset: 1380건
• 이미지의 크기: 1920x1080x4(png) 파일
• 주의 할 점: 한 이미지 내에 다양한 클래스가 포함 (max: 5)
• 총 label 값: 9가지
• all_dict
  • key: 한 이미지의 json 파일
  • value: class 값

train_data = '/mnt/dataset/train-valid_set'
inference_data = '/mnt/dataset/test_set'

import os
file_list = os.listdir(train_data)

file_list # 9개

['Suction Irrigator (SIRL-UCUK)',
 'R-Scissors (da Vinci) (MSCR-UCDV)',
 'Needle Holder (AESCULAP) (NDHL-UCAE)',
 'Graspers (straight) (OLYMPUS) (GRSL-UCOL)',
 'Metal Clip Applier (Medtronic) (CLAL-M3MT)',
 'R-Maryland (da Vinci) (BFCR-MADV)',
 'R-LND (da Vinci) (NDDR-LADV)',
 'Prograsp (da Vinci) (GRSR-UCDV)',
 'Polymer Clip Applier (CLAL-UCUK)']

class_list = ['Prograsp (da Vinci) (GRSR-UCDV)',
              'R-LND (da Vinci) (NDDR-LADV)',
              'R-Maryland (da Vinci) (BFCR-MADV)',
              'R-Scissors (da Vinci) (MSCR-UCDV)',
              'Suction Irrigator (SIRL-UCUK)',
              'Graspers (straight) (OLYMPUS) (GRSL-UCOL)',
              'Needle Holder (AESCULAP) (NDHL-UCAE)',
              'Metal Clip Applier (Medtronic) (CLAL-M3MT)',
              'Polymer Clip Applier (CLAL-UCUK)'
             ]

import os
data = os.listdir(train_data+'/'+file_list[0])
data.sort()
temp_data = data[:3] #한 사진에 json, png, xml 존재
temp_data, len(data)

(['GGHB_DC16_LWK0_LDG0_0015_0169548.json',
  'GGHB_DC16_LWK0_LDG0_0015_0169548.png',
  'GGHB_DC16_LWK0_LDG0_0015_0169548.xml'],
 4440)

# 주의!!!!!
# 수술도구마다 폴더가 나뉘어져 있어서 폴더내에서 접근하면 하나의 세그멘테이션만 가능.

# 그런데 하나의 사진에 여러 수술도구가 있음.

# 아래의 코드에서는
# Suction Irrigator (SIRL-UCUK),
# R-Scissors (da Vinci) (MSCR-UCDV)
# 위 두가지 도구가 같이 나오는 사진이 (306/3)=102장 인거임!!!

data2 = os.listdir(train_data+'/'+file_list[1])
len(list(set(data2).intersection(data)))

306

json_path = train_data+'/'+file_list[0]+'/'+temp_data[0]
xml_path = train_data+'/'+file_list[0]+'/'+temp_data[2]

#xml 파일 보기

def viewXML(xml_path):
    import xml.etree.ElementTree as elemTree

    tree = elemTree.parse(xml_path)
    root = tree.getroot()

    for child in root:
        print(child.tag, '   \t', child.text)
        for rechild in child:
            print('\t', rechild.tag+'\t', rechild.text)
            for points in rechild:
                print('\t\t', points.tag, points.text)
                
viewXML(xml_path)

folder    	 C:\Users\user\Desktop\0811~ 일\1111\Suction Irrigator (SIRL-UCUK)
filename    	 GGHB_DC16_LWK0_LDG0_0015_0169548.xml
path    	 C:\Users\user\Desktop\0811~ 일\1111\Suction Irrigator (SIRL-UCUK)
size    	 
    
	 width	 1920
	 height	 1080
	 depth	 3
object    	 
    
	 name	 Suction Irrigator (SIRL-UCUK)
	 id	 480
	 type	 poly
	 clr	 #8B6331
	 points	 
      
		 x 721.528662420382
		 y 566.296618233252
		 x 755.15923566879
		 y 570.879362221937
		 x 767.388535031847
		 y 580.044850199306
		 x 799.490445859873
		 y 592.265500835798
		 x 818.051881164882
		 y 605.577677767776
		 x 831.809842948321
		 y 628.491397711199
		 x 839.453155050232
		 y 650.533007586473
		 x 840.545280300412
		 y 677.810055291244
		 x 840.108743130212
		 y 703.77893789379
		 x 830.063694267516
		 y 738.474476019031
		 x 737.689925613962
		 y 800.449839269642
		 x 649.68152866242
		 y 855.009489520381
		 x 608.407643312102
		 y 893.199022759419
		 x 538.089171974522
		 y 946.664369294073
		 x 484.585987261147
		 y 984.853902533111
		 x 421.910828025478
		 y 1035.26408640864
		 x 382.165605095541
		 y 1073.45361964768
		 x 315.996583848907
		 y 1067.56271055677
		 x 264.901394493738
		 y 1012.58089237496
		 x 227.335728116097
		 y 917.645878873602
		 x 203.531152820607
		 y 867.230140156873
		 x 340.891719745223
		 y 790.851073678797
		 x 432.611464968153
		 y 738.913308473705
		 x 472.356687898089
		 y 717.527169859843
		 x 522.802547770701
		 y 685.447961939051
		 x 548.789808917197
		 y 668.644567313875
		 x 570.191082802548
		 y 656.423916677382
		 x 603.821656050955
		 y 635.037778063521
		 x 652.738853503185
		 y 602.958570142729

#json 파일 보기. 이거만 보면 될듯?!

import json

def getjson(json_path):
    with open(json_path) as json_file:
        json_data = json.load(json_file)
    return json_data

def viewjson(json_data):
    print(json.dumps(json_data, indent=4))
    
viewjson(getjson(json_path))

{
    "images": {
        "id": "GGHB_DC16_LWK0_LDG0_0015_0169548.png",
        "width": 1920,
        "height": 1080,
        "filename": "GGHB_DC16_LWK0_LDG0_0015_0169548.png",
        "date_created": "2021-11-11 18:06:35"
    },
    "metas": {
        "id": "\uc218\uc220\ub3c4\uad6c_\uc99d\ub840\uae30\ub85d\uc9c0_Suction Irrigator (SIRL-UCUK).xlsx",
        "operation": "LDG",
        "hospital": "\uac00\ucc9c\ub300\uae38\ubcd1\uc6d0"
    },
    "annotations": [
        {
            "id": "GGHB_DC16_LWK0_LDG0_0015_0169548.json",
            "image_id": "GGHB_DC16_LWK0_LDG0_0015_0169548.png",
            "category_id": 480,
            "category_name": "Suction Irrigator (SIRL-UCUK)",
            "types": "poly",
            "points": [
                [
                    721.528662420382,
                    566.296618233252
                ],
                [
                    755.15923566879,
                    570.879362221937
                ],
                [
                    767.388535031847,
                    580.044850199306
                ],
                [
                    799.490445859873,
                    592.265500835798
                ],
                [
                    818.051881164882,
                    605.577677767776
                ],
                [
                    831.809842948321,
                    628.491397711199
                ],
                [
                    839.453155050232,
                    650.533007586473
                ],
                [
                    840.545280300412,
                    677.810055291244
                ],
                [
                    840.108743130212,
                    703.77893789379
                ],
                [
                    830.063694267516,
                    738.474476019031
                ],
                [
                    737.689925613962,
                    800.449839269642
                ],
                [
                    649.68152866242,
                    855.009489520381
                ],
                [
                    608.407643312102,
                    893.199022759419
                ],
                [
                    538.089171974522,
                    946.664369294073
                ],
                [
                    484.585987261147,
                    984.853902533111
                ],
                [
                    421.910828025478,
                    1035.26408640864
                ],
                [
                    382.165605095541,
                    1073.45361964768
                ],
                [
                    315.996583848907,
                    1067.56271055677
                ],
                [
                    264.901394493738,
                    1012.58089237496
                ],
                [
                    227.335728116097,
                    917.645878873602
                ],
                [
                    203.531152820607,
                    867.230140156873
                ],
                [
                    340.891719745223,
                    790.851073678797
                ],
                [
                    432.611464968153,
                    738.913308473705
                ],
                [
                    472.356687898089,
                    717.527169859843
                ],
                [
                    522.802547770701,
                    685.447961939051
                ],
                [
                    548.789808917197,
                    668.644567313875
                ],
                [
                    570.191082802548,
                    656.423916677382
                ],
                [
                    603.821656050955,
                    635.037778063521
                ],
                [
                    652.738853503185,
                    602.958570142729
                ]
            ]
        }
    ]
}

import matplotlib.pyplot as plt
from PIL import Image
import numpy as np

# json_file을 열어서 어노테이션 입힌 이미지 보기

def viewImage(json_path):
    json_data = getjson(json_path)
    
    image = Image.open(json_path.rpartition('/')[0]+'/'+json_data["images"]["filename"])
    plt.imshow(image)
    color_list = ["blue","yellow","red","green","pink","black"]
    for i in range(len(json_data["annotations"])):
        points = json_data["annotations"][i]["points"].copy()
        points.append(points[0])
        points = np.array(points)
        #plt.plot(points[:,0], points[:,1], linestyle='solid')
        plt.fill(points[:,0], points[:,1], color=color_list[i], alpha=0.5)
    plt.show()
    
viewImage(json_path)

# 수술도구마다 하나씩 보기!

for tool in file_list:
    print(tool)
    data = os.listdir(train_data+'/'+tool)
    data.sort()

    for file_name in data:
        if(file_name[-4:] == 'json'):
            print(file_name)
            json_path = train_data+'/'+tool+'/'+file_name
            viewImage(json_path)
            break

Suction Irrigator (SIRL-UCUK)
GGHB_DC16_LWK0_LDG0_0015_0169548.json

R-Scissors (da Vinci) (MSCR-UCDV)
SNUH_DC07_JCW0_RALP_0000_00005250.json

Needle Holder (AESCULAP) (NDHL-UCAE)
SNUH_DC07_JCW0_RALP_0000_00182178.json

Graspers (straight) (OLYMPUS) (GRSL-UCOL)
SNUH_DC07_JCW0_RALP_0000_00151399.json

Metal Clip Applier (Medtronic) (CLAL-M3MT)
SNUH_DC07_JCW0_RALP_0001_00046340.json

R-Maryland (da Vinci) (BFCR-MADV)
SNUH_DC07_JCW0_RALP_0000_00004758.json

R-LND (da Vinci) (NDDR-LADV)
SNUH_DC07_JCW0_RALP_0000_00181898.json

Prograsp (da Vinci) (GRSR-UCDV)
GGHB_DC68_LJH0_BABA_0002_0000060.json

Polymer Clip Applier (CLAL-UCUK)
SNUH_DC07_JCW0_RALP_0001_00044763.json

# 근데 이미지 하나당 수술도구 하나씩 나오니까
# 처음부터 데이터를 전부 읽어와서 리스트 하나로 만들어봤어

# all_data 는 모든 json file 리스트
# data_dic 는 key 값이 수술도구, value가 json file인 딕셔너리

all_data = []
data_dic = {}

for tool in file_list:
    files = os.listdir(train_data+'/'+tool)
    json_files = [file for file in files if file.endswith(".json")]
    
    all_data.extend(json_files)
    data_dic[tool] = json_files

len(all_data) # 모든 json file은 12408개

12408

# 중복을 제거하니, 수술도구가 여러개 있는 사진이 2천장정도 있음

all_data = list(set(all_data))
len(all_data)

10792

# all_dic 은 key 값은 사진, value값은 수술도구인 딕셔너리.
# 만들기 어려웠음! #수고했음!
# 이렇게 하면 사진 하나에서 여러 어노테이션을 동시에 볼 수 있음.

all_dic = {}
        
for files in all_data:
    for i in data_dic:
        if(files in data_dic[i]):
            if(files in all_dic):
                temp = all_dic[files]
                temp.append(i)
                all_dic[files] = temp
            else:
                all_dic[files] = [i]
all_dic

{'SNUH_DC16_KSH0_LDG0_0050_00292440.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC08_KHS0_ENME_0039_00064081.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_KMJ0_RHC0_0001_00067272.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0031_00041643.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0021_00131778.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0009_0010499.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0011_00038969.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0011_0069255.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Graspers (straight) (OLYMPUS) (GRSL-UCOL)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0001_00012091.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC08_KHS0_ENME_0016_00026322.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0001_0015934.json': ['Suction Irrigator (SIRL-UCUK)',
  'R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0020_0011157.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0004_00051872.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00071460.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0032_00036741.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0021_0000533.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC16_KSH0_LDG0_0046_00273852.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0000_00009491.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0031_00092520.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0001_00005348.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0003_0024449.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0016_00019918.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0020_00065117.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0034_00218466.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0011_0022546.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0014_0043664.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0006_00027208.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0031_00037665.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0039_0061560.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0031_0054144.json': ['R-Maryland (da Vinci) (BFCR-MADV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00169995.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC16_PJW0_LAR0_0010_0017154.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0004_00038343.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0002_0075040.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0032_00015252.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_00012249.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0031_00035590.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00089911.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0031_0067996.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0008_00046272.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0002_00035010.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0009_0005628.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC68_LKE0_BABA_0002_0302126.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'GGHB_DC68_LJH0_BABA_0003_0025146.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0007_00026115.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC08_KHS0_ENME_0039_00063836.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00107732.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RLPN_0003_0043651.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0039_00068349.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0035_00393662.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0031_0063874.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0007_00026055.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0020_00037816.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0007_00056420.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RLPN_0002_0079266.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0008_00042508.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0016_00043871.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0032_00019933.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0031_0055651.json': ['Suction Irrigator (SIRL-UCUK)',
  'R-Maryland (da Vinci) (BFCR-MADV)',
  'Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00182211.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0022_00082530.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC08_KHS0_ENME_0007_00119428.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0016_00063029.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0032_00016770.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0031_0042958.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC16_KSH0_LDG0_0035_00437486.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RLPN_0003_0016706.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC16_PJW0_LAR0_0010_0017461.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0011_0086920.json': ['Suction Irrigator (SIRL-UCUK)',
  'Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC08_KHS0_ENME_0015_00029878.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0008_00037777.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0000_0018817.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC68_KSJ0_BABA_0001_0244464.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0031_0018709.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'GGHB_DC68_LJH0_BABA_0003_0005940.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0016_00024441.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0003_0044791.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0003_0036331.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0013_00138811.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0002_00039520.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0039_00064102.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0031_00075999.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0000_00008627.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_00063919.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0000_00138929.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0031_0043609.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_00058731.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0009_0079355.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0032_00023914.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0031_0041397.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)',
  'Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00198610.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'GGHB_DC68_LJH0_BABA_0003_0015141.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'GGHB_DC68_LJH0_BABA_0003_0019161.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0008_0083539.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0000_00091184.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'GGHB_DC68_LJH0_BABA_0003_0024006.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0020_00107795.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0031_00028029.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0011_00074295.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0008_00082280.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0021_00017897.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0094865.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0062911.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0002_00022261.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC16_KSH0_LDG0_0045_00369164.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0009_00038755.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC08_KHS0_ENME_0032_00021301.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0039_00067043.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00077060.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0011_00074276.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0022_00082640.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0011_00074488.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_0058385.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0047061.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0005_00050942.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0021_00063962.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC08_KHS0_ENME_0032_00023264.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0008_0083599.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0022_00082529.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0007_00057743.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC08_KHS0_ENME_0039_00070842.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0055_00427676.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC16_KSH0_LDG0_0058_00128516.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RLPN_0003_0016106.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_00064500.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0004_00055337.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC08_KHS0_ENME_0039_00067440.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0056_00277582.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RLPN_0002_0074250.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0006_00043154.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00093065.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'GGHB_DC68_LJH0_BABA_0002_0020903.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0032_00018492.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00074267.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0009_0075057.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0037_00244959.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC08_KHS0_ENME_0039_00069426.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0014_0018469.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0000_0007517.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0020_00065419.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0020_0000521.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0001_0049678.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0031_0065996.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC08_KHS0_ENME_0032_00023894.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_00058541.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0009_0067171.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0004_00051023.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0021_00059530.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0003_0064341.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0031_0131536.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0021_00002961.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00077065.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0014_0019893.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0003_0045331.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0008_0035117.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_00180850.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0011_00070610.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0007_00057604.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0011_0071383.json': ['Prograsp (da Vinci) (GRSR-UCDV)',
  'Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0004_00013342.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC16_KSH0_LDG0_0037_00221101.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RLPN_0014_0015604.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_00012067.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0022_00158941.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC16_KSH0_LDG0_0056_00021252.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC16_RSB0_RHC0_0005_00047517.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0001_00056973.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0002_00007009.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0016_00024572.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0006_00027096.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0011_00032838.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0000_00197349.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RLPN_0008_0035197.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0014_0029341.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0000_00165310.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC08_KHS0_ENME_0039_00183792.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0037_00217536.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0013_00082402.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0008_00078243.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC08_KHS0_ENME_0039_00068546.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0020_0073595.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0001_00012001.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC08_KHS0_ENME_0032_00038056.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0002_00020505.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00021645.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0048_00191173.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC16_RSB0_RHC0_0005_00047513.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0001_0088322.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RLPN_0002_0062891.json': ['R-Maryland (da Vinci) (BFCR-MADV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00207777.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RLPN_0003_0017426.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0049_00288376.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC08_KHS0_ENME_0039_00065382.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_0040262.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC08_KHS0_ENME_0039_0336030.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0020_00090818.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0032_00023897.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0008_00010997.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC08_KHS0_ENME_0012_00021000.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0011_0048718.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0021_00004743.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_0037473.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0011_0057768.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_00052500.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0004_00013294.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0021_0000603.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0020_00064576.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0037_00221010.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC08_KHS0_ENME_0019_00060931.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0022_00082600.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RLPN_0014_0004763.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0014_0007973.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_0074300.json': ['Prograsp (da Vinci) (GRSR-UCDV)',
  'Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC68_LKE0_BABA_0002_0170659.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0035_00223845.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0039_00068577.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0035_00315781.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0011_00107735.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0000_00010282.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0031_0055173.json': ['Suction Irrigator (SIRL-UCUK)',
  'R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0047_00165131.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0021_00008345.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00078403.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'GGHB_DC68_LJH0_BABA_0003_0027366.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0031_0030624.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0008_0085639.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00007484.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0032_00023511.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0002_00021578.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_0000974.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0039_00184459.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0023_00029727.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0032_00021774.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0039_00070777.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0023_00030944.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0003_0093452.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC08_KHS0_ENME_0020_0008432.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0016_00060080.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0011_00068770.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'GGHB_DC68_LJH0_BABA_0003_0025026.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00074518.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0003_0099322.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC16_KMJ0_RHC0_0001_00050849.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC08_KHS0_ENME_0012_00008126.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0001_00088090.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC08_KHS0_ENME_0039_00070574.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00069782.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0006_00019936.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0039_00068523.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00077760.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0048_00183236.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC16_KSH0_LDG0_0043_00153775.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0021_00062245.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0008_0085399.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_PJW0_LAR0_0008_0019881.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC16_KSH0_LDG0_0035_00318041.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0011_00091308.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC08_KHS0_ENME_0039_00063832.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0031_0363022.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0020_00061258.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0039_00183931.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0008_00082399.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0000_0010266.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0001_00065739.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0009_0010437.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0001_00063591.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC16_KSH0_LDG0_0053_00096510.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0009_0002148.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0009_0015881.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0011_0076036.json': ['Suction Irrigator (SIRL-UCUK)',
  'R-Maryland (da Vinci) (BFCR-MADV)',
  'Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC16_KSH0_LDG0_0058_00143863.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC08_KHS0_ENME_0011_00025098.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC16_KMJ0_RHC0_0001_00067007.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0020_00062145.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0031_0055668.json': ['Suction Irrigator (SIRL-UCUK)',
  'R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)',
  'Prograsp (da Vinci) (GRSR-UCDV)',
  'Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC16_KSH0_LDG0_0041_00075440.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC16_KSH0_LDG0_0037_00206373.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC08_KHS0_ENME_0039_00068729.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0020_0074398.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00196034.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RLPN_0003_0093483.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC08_KHS0_ENME_0039_00069554.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0133001.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0006_00027633.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0011_00065589.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0001_0049249.json': ['Suction Irrigator (SIRL-UCUK)',
  'R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_0059218.json': ['Suction Irrigator (SIRL-UCUK)',
  'Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0021_00065055.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC16_KSH0_LDG0_0035_00320200.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0001_00047385.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'GGHB_DC68_LJH0_BABA_0002_0015061.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_00002147.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00076581.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00167832.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC16_KSH0_LDG0_0031_0003872.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC68_KSJ0_BABA_0001_0226385.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0020_00094456.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0039_00070185.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_00092751.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0056_00275395.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0031_00047501.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0039_00068576.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0003_0030451.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0011_00075680.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0001_0094370.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0011_00075079.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0002_00012160.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0039_00068038.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0007_00030680.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0002_00020514.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_0061659.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0011_00107698.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0000_00055038.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_PJW0_LAR0_0010_0016787.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0016_00026774.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0020_0010879.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0012_00022242.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0011_00051460.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00040351.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0011_00069310.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_0013147.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0031_0023661.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0031_00018634.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0003_0030571.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0014_0024125.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_0049392.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_0015845.json': ['Suction Irrigator (SIRL-UCUK)',
  'R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0021_00052258.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC16_KSH0_LDG0_0054_00187694.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0016_00019956.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0021_00018167.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0032_00036517.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0038_00202296.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC16_KSH0_LDG0_0051_00197908.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RLPN_0014_0018494.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0001_0072006.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0021_00004653.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0005_00070787.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC08_KHS0_ENME_0016_00060024.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC16_RSB0_RHC0_0010_00136438.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC08_KHS0_ENME_0020_0073638.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0003_0093303.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0021_00052144.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_00198373.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC08_KHS0_ENME_0032_00018184.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0032_00037925.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC16_KSH0_LDG0_0036_00243170.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0007_00030135.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0000_0014227.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00050314.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0003_0058002.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00166815.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0011_00050430.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00074278.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0032_00016719.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0023_00029680.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0004_00041480.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0003_0024629.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0003_0059862.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0008_00030257.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0009_0023880.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC68_KSJ0_BABA_0001_0088279.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_00131796.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0021_00202907.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0021_00067519.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0031_0030833.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC16_KSH0_LDG0_0038_00191614.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC08_KHS0_ENME_0013_00009050.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0032_00021805.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0046911.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC16_KMJ0_RHC0_0001_00050912.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0001_00012009.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0013_00087320.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0031_0052356.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0000_00015931.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0000_00026468.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0004_00056750.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0007_00026517.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC08_KHS0_ENME_0032_0018760.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0016_00018504.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0008_0025536.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0020_00094759.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'GGHB_DC68_LJH0_BABA_0002_0013615.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0008_0025296.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_00057003.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC16_KSH0_LDG0_0036_00135192.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0002_00012596.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0080069.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0008_0025356.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0032_0243740.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0031_0054142.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0006_00027058.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC16_KSH0_LDG0_0037_00222635.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC08_KHS0_ENME_0008_00041635.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00026654.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0031_00035734.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0032_00019053.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0031_00044532.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0032_00021948.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0011_00077573.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0046841.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC16_KSH0_LDG0_0036_00130063.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0031_00037530.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00019825.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0011_00074271.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0002_00012565.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_0046377.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0000_00015539.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0050078.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0000_00221827.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0011_00074248.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0031_00089824.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RLPN_0008_0080599.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_00062515.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00182225.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC16_KSH0_LDG0_0035_00421505.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0006_00027081.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC16_KSH0_LDG0_0040_00313051.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0021_00006814.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0023_00029671.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00106791.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0061181.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00060839.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0006_00026005.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00105560.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0000_00167961.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0021_00132969.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0039_0068220.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0002_00043223.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00160307.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0000_00008764.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0012_00008692.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC16_KSH0_LDG0_0050_00291270.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0031_0050486.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0003_0092530.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0000_0019126.json': ['Suction Irrigator (SIRL-UCUK)',
  'R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0039_00070820.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00165352.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0000_00016928.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_0110421.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'GGHB_DC68_LJH0_BABA_0003_0018861.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0031_0055592.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0003_0074840.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC16_KSH0_LDG0_0054_00187150.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0021_0117062.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0006_00027139.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC16_PJW0_LAR0_0010_0016806.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0015_00029428.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0031_00046697.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0009_0011304.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0001_00056502.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0020_0008532.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0009_0006996.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_00011996.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0005_00051199.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC16_KSH0_LDG0_0031_0360864.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC08_KHS0_ENME_0011_00075590.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0039_00186141.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_0062028.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0001_0074942.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'GGHB_DC68_LJH0_BABA_0003_0026226.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0034_00220180.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0011_00070203.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00007178.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0001_00056953.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0013_00009177.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00064254.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_0107838.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC16_PJW0_RHC0_0007_00037155.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0020_00100505.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0002_00034471.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RLPN_0003_0022347.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0035_00430443.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC08_KHS0_ENME_0023_00029659.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0039_00100276.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00058510.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0021_00059721.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00058490.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0011_00077084.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0002_0061261.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0037_00217986.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0001_00009787.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'GGHB_DC68_LJH0_BABA_0003_0006002.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0054_00225225.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0001_00049296.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0008_00011003.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RLPN_0014_0003096.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_0083273.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0031_00046310.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00168574.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0039_00064130.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00119546.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0039_00069506.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0054_00022835.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC16_KSH0_LDG0_0036_00034250.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC16_KSH0_LDG0_0058_00142311.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC16_KSH0_LDG0_0031_0131288.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0017_00021418.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0014_0019013.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0061221.json': ['R-Maryland (da Vinci) (BFCR-MADV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0014_0020138.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0012_00008678.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0008_00050721.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0001_00056959.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC16_KSH0_LDG0_0047_00164051.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC16_KSH0_LDG0_0047_00168175.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC16_RSB0_LAR0_0002_00056053.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0039_0327450.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC16_KSH0_LDG0_0047_00164264.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0020_00096774.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0011_00069821.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0002_0061581.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0040_00336659.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC08_KHS0_ENME_0013_00019215.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00167883.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC08_KHS0_ENME_0032_0183563.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0009_0006685.json': ['R-Maryland (da Vinci) (BFCR-MADV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0020_00060403.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0002_00012060.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00012531.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0020_00071464.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0008_00017607.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0001_0102010.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0020_00064276.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0039_0081240.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0032_00018493.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0009_0014333.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0000_0009044.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0047722.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0020_0002665.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0032_0024454.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00035417.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC16_KSH0_LDG0_0035_00223170.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0022_00158974.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0031_00030074.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0017_00022122.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0011_00027138.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00017762.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00206315.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0127453.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC16_KSH0_LDG0_0035_00437576.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'GGHB_DC68_LJH0_BABA_0002_0025226.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0008_0086719.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00171365.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0032_00037798.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_00119521.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC08_KHS0_ENME_0016_00024051.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0020_00041589.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0001_0014489.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00039796.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0011_00077062.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0008_0020605.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0037_00206193.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0001_0072639.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0012_00058210.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0011_0069256.json': ['Suction Irrigator (SIRL-UCUK)',
  'Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC08_KHS0_ENME_0020_0009367.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC16_PJW0_LAR0_0008_0019866.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00151664.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0000_0019621.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0001_00009132.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC16_KSH0_LDG0_0035_00394473.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0031_0019942.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0002_00034409.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0000_00206266.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0020_00089556.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_0016836.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0014_0020165.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0039_00064105.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_00131840.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0005_00050928.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RLPN_0014_0018623.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0039_00186090.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'GGHB_DC16_LWK0_LDG0_0017_0228913.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0011_00102566.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0000_0009838.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0021_00265848.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RLPN_0003_0093429.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0002_00034523.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC16_KSH0_LDG0_0035_00223520.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC68_LKE0_BABA_0001_0123279.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0074590.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_00071841.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RLPN_0009_0055016.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0021_00131730.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC16_KSH0_LDG0_0048_00187802.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC08_KHS0_ENME_0039_00068574.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_RSB0_LAR0_0004_00020721.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0021_00063956.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0021_0087480.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0032_00037666.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_RSB0_RHC0_0003_00040860.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0000_00012349.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0002_00038828.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC08_KHS0_ENME_0032_0080751.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0021_00032713.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC08_KHS0_ENME_0011_00069500.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC16_KSH0_LDG0_0035_00424754.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC16_KSH0_LDG0_0054_00188035.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC16_PJW0_LAR0_0010_0016785.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0001_0071881.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)',
  'R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0000_00017206.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0008_00010991.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC08_KHS0_ENME_0039_00067442.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0039_00064164.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_RSB0_RHC0_0010_00136692.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0031_00071178.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC68_LKE0_BABA_0002_0244828.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00192649.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0031_00089834.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0001_0050949.json': ['R-Maryland (da Vinci) (BFCR-MADV)',
  'Prograsp (da Vinci) (GRSR-UCDV)',
  'Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0014_0007772.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0039_00183825.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00071310.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0004_00051915.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0010_00042374.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0021_00094000.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_0022252.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00059318.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC16_RSB0_RHC0_0010_00136649.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0011_0050507.json': ['R-Maryland (da Vinci) (BFCR-MADV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_0106113.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0009_00038869.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0008_00011098.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0000_00165412.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0013_00087334.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC08_KHS0_ENME_0039_00184002.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00095295.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0097544.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0011_00066915.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0007_00030201.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC08_KHS0_ENME_0020_0058110.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0032_0026048.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0014_0003240.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_00088082.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0011_00069813.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0011_00075105.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'GGHB_DC68_LJH0_BABA_0003_0006805.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0008_00078202.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC16_KSH0_LDG0_0038_00007032.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC08_KHS0_ENME_0032_00018464.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0006_00023575.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0019_00020516.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0032_00018156.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00015749.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0074960.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0031_00057427.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0002_00034388.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC08_KHS0_ENME_0007_00075135.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00071155.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0001_0010546.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0021_00132996.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00184903.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0052238.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0020_00066744.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC16_PJW0_LAR0_0008_0019878.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0039_00187879.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00035483.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0011_00087081.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0011_0027588.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0000_00014291.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'GGHB_DC68_LJH0_BABA_0003_0026346.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_00155206.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0021_00109362.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC08_KHS0_ENME_0039_00064145.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0039_00179323.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0003_0099802.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0031_00038258.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0021_00082376.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0003_0099682.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0020_0011507.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0001_00071861.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RLPN_0008_0022225.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_0089284.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0011_00066908.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC08_KHS0_ENME_0032_00021890.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0011_00062297.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00222046.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0007_00056445.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0008_00082408.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC08_KHS0_ENME_0039_00186107.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0032_00023526.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0035_00429810.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RLPN_0009_0073383.json': ['R-Maryland (da Vinci) (BFCR-MADV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0002_00034503.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC08_KHS0_ENME_0039_00064079.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0020_00107374.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0000_00138505.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0020_00062265.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0097534.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0006_00021377.json': ['Suction Irrigator (SIRL-UCUK)'],
 'GGHB_DC68_LJH0_BABA_0003_0012048.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00066864.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC16_KSH0_LDG0_0043_00214035.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RLPN_0009_0004259.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0032_00021642.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0020_00102310.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0031_0019578.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0009_00038631.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC08_KHS0_ENME_0039_00063962.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_00059903.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0006_00020064.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0008_00046289.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC16_KMJ0_RHC0_0001_00050806.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0001_00012078.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC16_KSH0_LDG0_0053_00249004.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0011_00081165.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0021_00004473.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_0197981.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RLPN_0014_0014357.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0039_00061404.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0053_00096860.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0031_00098915.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC08_KHS0_ENME_0032_00016715.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0019_00014270.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0009_0030744.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0020_00071524.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0009_0000373.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0016_00026384.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0002_0046941.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0000_00198012.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RLPN_0008_0083419.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_0050944.json': ['R-Maryland (da Vinci) (BFCR-MADV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC68_LKE0_BABA_0002_0265812.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0031_0068545.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0000_00035272.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00196597.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0021_00010146.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0126793.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RLPN_0008_0035357.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0047692.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0011_0104141.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0000_00059463.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_RSB0_RHC0_0003_00040825.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0000_00057074.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00167965.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0032_0025865.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0022_00082620.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0009_00038706.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0001_00088085.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0008_00037678.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC68_LKE0_BABA_0002_0307428.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0049888.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0031_00099200.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0000_00166999.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC16_KSH0_LDG0_0047_00167836.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC16_KSH0_LDG0_0051_00115790.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0020_0057591.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0021_00022611.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0016_00029015.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0011_00032357.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0031_00076020.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0000_0015410.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00052250.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_0009471.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0000_00047282.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0007_00026101.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0021_00132951.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0001_00012407.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC16_KSH0_LDG0_0056_00275304.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RLPN_0014_0015331.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0002_00021425.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_0035304.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0012_00052934.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC08_KHS0_ENME_0023_00029795.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC16_RSB0_LAR0_0004_00023250.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0021_00155196.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0000_00190879.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0011_00069809.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0001_00011225.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0045730.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0032_00021763.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0020_0001265.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00202549.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC16_PJW0_RHC0_0001_00119133.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC08_KHS0_ENME_0039_00183640.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0031_0284081.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC16_KSH0_LDG0_0051_00204120.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'GGHB_DC68_LJH0_BABA_0002_0012590.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0031_0028141.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0031_0019698.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0031_00020509.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0006_00020091.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0039_00064064.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0006_00043278.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0011_00070201.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC16_KSH0_LDG0_0036_00240370.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0021_00014020.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_PJW0_RHC0_0001_00120522.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'GGHB_DC68_LJH0_BABA_0003_0001552.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00066612.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00221575.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0011_00069784.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0001_0000560.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0008_0023228.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0039_00187613.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_00064563.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RLPN_0003_0028651.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0003_0039451.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0021_00050524.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0012861.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0002_00007014.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0039_00064222.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0040_00279847.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RLPN_0003_0042151.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC16_KSH0_LDG0_0040_00337293.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC16_KSH0_LDG0_0048_00116155.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0039_00069422.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_00045617.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0020_0056013.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0008_0035437.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0039_00183729.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00093246.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0019_00001049.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00151515.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC08_KHS0_ENME_0039_00068539.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0032_00021768.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_0056907.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0031_0046373.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_0008311.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0003_0102745.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC08_KHS0_ENME_0020_0009416.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0012_00008696.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0021_00059014.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00133476.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0039_00068507.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0039_00070550.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00024244.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0009_0003949.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0049_00013544.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0021_0057280.json': ['Suction Irrigator (SIRL-UCUK)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0014_0035270.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0000_00155062.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0011_00078463.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0039_00066764.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_0010694.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_00250197.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC16_KSH0_LDG0_0033_0078090.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC08_KHS0_ENME_0032_00019918.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'GGHB_DC68_LJH0_BABA_0003_0006064.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0035_00223510.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC16_KSH0_LDG0_0048_00183692.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0008_00082291.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RLPN_0002_0127328.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0000_00044780.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0001_0053517.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0011_0057928.json': ['R-Maryland (da Vinci) (BFCR-MADV)',
  'Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC16_KSH0_LDG0_0037_00223179.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0013_00082262.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC08_KHS0_ENME_0032_00019126.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0031_0080826.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC08_KHS0_ENME_0039_00063824.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0032_00016401.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0002_00066240.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0001_00108505.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC68_LKE0_BABA_0002_0311024.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0031_0018222.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC68_LKE0_BABA_0002_0211592.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_0091980.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC16_RSB0_RHC0_0003_00085358.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0005_00070770.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0001_00059125.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0003_0057462.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0004_00051877.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0003_0026851.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0004_00053978.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC08_KHS0_ENME_0032_00021889.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0057_00122042.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0032_00021762.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00009185.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'GGHB_DC68_LJH0_BABA_0003_0024246.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0031_0074726.json': ['Suction Irrigator (SIRL-UCUK)',
  'R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0002_00037461.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0039_00068473.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00061761.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0003_0053737.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0015_00029906.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0011_00070211.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0039_00186251.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0009_0005006.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0004_00052114.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0031_0055904.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_0008861.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0021_0001037.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC16_KSH0_LDG0_0035_00312876.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RLPN_0002_0090640.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00052000.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0039_00068527.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00074297.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0001_00108528.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RLPN_0008_0038249.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0032_00018158.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0002_00034405.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0002_00006753.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0020_0038872.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC68_LKE0_BABA_0002_0387445.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0054_00187671.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0032_00018169.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0031_0020750.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0004_00030217.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0001_00048030.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_00032184.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0032_00018304.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_KSH0_LDG0_0040_00309332.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC08_KHS0_ENME_0032_0016831.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0013411.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0001_0072099.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0001_0054797.json': ['Suction Irrigator (SIRL-UCUK)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0039_00100280.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0003_0018206.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0020_00060769.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00042695.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0016_00024694.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0032_00017819.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_0050923.json': ['R-Maryland (da Vinci) (BFCR-MADV)',
  'Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_0011488.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC16_PJW0_RHC0_0007_00152710.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0031_00098571.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0011_0095504.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0001_0000506.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0009_0011118.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0000_00139596.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0008_00046590.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0021_00005734.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0004_00051705.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0022_00089631.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0002_00114319.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RLPN_0008_0087559.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0011_0009737.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0000_0017269.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0016_00042748.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0009_0000062.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'GGHB_DC68_LJH0_BABA_0003_0016461.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00066905.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0001_0081836.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)',
  'Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0021_00265858.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC16_KSH0_LDG0_0035_00426829.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0001_00050250.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC16_RSB0_LAR0_0002_00082627.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0031_0029777.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_0016909.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0007_00068767.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0002_00026279.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0003_0064362.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0000_00166019.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0020_00099237.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0004_00054115.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0007_00026583.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0000_00106514.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0007_00026009.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC07_JCW0_RALP_0022_00089630.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'SNUH_DC08_KHS0_ENME_0039_00186080.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_00012007.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0021_0001008.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RLPN_0014_0002810.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0000_00106883.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0020_00099964.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0021_00131790.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0014_0023421.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0031_00092537.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC68_LKE0_BABA_0002_0273250.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0001_0016204.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0031_0019897.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0039_00183987.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RLPN_0002_0107513.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC08_KHS0_ENME_0020_0000771.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0001_00004925.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0004_00030226.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0003_0020546.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0039_00186296.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0011_00031455.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0000_00219012.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0021_00052289.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00174825.json': ['Graspers (straight) (OLYMPUS) (GRSL-UCOL)'],
 'SNUH_DC07_JCW0_RALP_0000_00023100.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC07_JCW0_RALP_0001_00000000.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_00084922.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC16_KMJ0_RHC0_0001_00066984.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'GGHB_DC68_LJH0_BABA_0003_0023526.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0031_00043991.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0032_00021892.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0006_00022699.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0002_0133051.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0011_00070209.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RLPN_0002_0094835.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_00133114.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC16_KSH0_LDG0_0033_0006090.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RLPN_0009_0074996.json': ['R-Scissors (da Vinci) (MSCR-UCDV)',
  'R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC08_KHS0_ENME_0020_0006015.json': ['Suction Irrigator (SIRL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0011_0104228.json': ['R-LND (da Vinci) (NDDR-LADV)'],
 'SNUH_DC07_JCW0_RALP_0009_00038725.json': ['Metal Clip Applier (Medtronic) (CLAL-M3MT)'],
 'GGHB_DC68_LJH0_BABA_0003_0023106.json': ['Prograsp (da Vinci) (GRSR-UCDV)'],
 'SNUH_DC08_KHS0_ENME_0032_00017917.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0031_00043508.json': ['R-Maryland (da Vinci) (BFCR-MADV)'],
 'SNUH_DC16_PJW0_LAR0_0010_0019610.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0000_00193671.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 'SNUH_DC07_JCW0_RALP_0001_00045590.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC07_JCW0_RALP_0021_00131784.json': ['Polymer Clip Applier (CLAL-UCUK)'],
 'SNUH_DC08_KHS0_ENME_0039_00077836.json': ['R-Scissors (da Vinci) (MSCR-UCDV)'],
 'SNUH_DC07_JCW0_RALP_0021_00202920.json': ['Needle Holder (AESCULAP) (NDHL-UCAE)'],
 ...}

data_filenames = list(all_dic.keys())
data_filenames.sort()

len(data_filenames)

10792

maximum = 0
for i in all_dic:
    if maximum < len(all_dic[i]):
        maximum = len(all_dic[i])
maximum

5

import matplotlib.pyplot as plt
from PIL import Image
import numpy as np
    
# image_file을 열어서 어노테이션 입힌 여러 이미지 보기
    
def plotImage(json_file):
    tools = all_dic[json_file]
    image_path = train_data+'/'+tools[0]+'/'+json_file.rpartition('.')[0]+'.png'
    
    image = Image.open(image_path)
    plt.figure(figsize=(3, 2))
    plt.imshow(image)
    color_list = ["blue","yellow","red","green","pink"]
    
    for tool in range(len(tools)):
        json_path = train_data+'/'+tools[tool]+'/'+json_file
        json_data = getjson(json_path)
    
        for i in range(len(json_data["annotations"])):
            points = json_data["annotations"][i]["points"].copy()
            points.append(points[0])
            points = np.array(points)
            #plt.plot(points[:,0], points[:,1], linestyle='solid')
            plt.fill(points[:,0], points[:,1], color=color_list[tool], alpha=0.5)
    plt.show()
    
    
plotImage('SNUH_DC07_JCW0_RLPN_0008_0023168.json')

#어노테이션 제일 많은(5개)인 애들 시각화.

for i in all_dic:
    if maximum == len(all_dic[i]):
        plotImage(i)

y label masking

from PIL import Image, ImageDraw

def getData(file):
    tools = all_dic[file]

    json_path = train_data + '/' + tools[0] + '/' + file
    json_data = getjson(json_path)
    width = json_data['images']['width']
    height = json_data['images']['height']
    
    image_path = train_data+'/'+tools[0]+'/'+ file.rpartition('.')[0]+'.png'
#    image = np.array(Image.open(image_path))
    
#    file_name = json_data['images']['filename'].rpartition('.')[0]
    mask = np.zeros((height,width,9))
    for tool in tools:
        class_num = class_list.index(tool)
        json_path = train_data + '/' + tool + '/' + file
        json_data = getjson(json_path)
        for classes in class_list:
            for json_seg in json_data['annotations']:
                if(classes == json_seg['category_name']):
                    found = True
                    points = json_seg["points"].copy()
                    points = [(x[0],x[1]) for x in points]
                    img = Image.new('L', (width, height), 0)
                    ImageDraw.Draw(img).polygon(points, outline=1, fill=1)
                    mask[:,:,class_num] = np.array(img)
    return image_path, mask

for i in all_dic:
    x = getData(i)
    break

# y label 만들기

import numpy as np

for file in all_dic:
    image, mask = getData(file)
#    with open('labels/'+file_name, 'wb') as f:
#        np.save(f, mask)
    print(file)
    for i in range(9):
        plt.imshow(mask[:,:,i])
        plt.show()
    break
        

SNUH_DC16_KSH0_LDG0_0050_00292440.json

DataLoader

from sklearn.model_selection import train_test_split
import tensorflow as tf
import tensorflow_addons as tfa

'''
wset = set()
itr = 0
for i in train_filenames:
    wset.add((getData(i)[0].shape[0], getData(i)[0].shape[1]))
    if(len(wset) != itr):
        itr = len(wset)
        print(wset)
print(wset)
'''

'\nwset = set()\nitr = 0\nfor i in train_filenames:\n    wset.add((getData(i)[0].shape[0], getData(i)[0].shape[1]))\n    if(len(wset) != itr):\n        itr = len(wset)\n        print(wset)\nprint(wset)\n'

def shift_img(output_img, label_img, width_shift_range, height_shift_range):
    """This fn will perform the horizontal or vertical shift"""
    if width_shift_range or height_shift_range:
        if width_shift_range:
            width_shift_range = tf.random.uniform([],  
                                                  -width_shift_range * img_shape[1],
                                                  width_shift_range * img_shape[1])
        if height_shift_range:
            height_shift_range = tf.random.uniform([],
                                                   -height_shift_range * img_shape[0],
                                                   height_shift_range * img_shape[0])
        output_img = tfa.image.translate(output_img,
                                         [width_shift_range, height_shift_range])
        label_img = tfa.image.translate(label_img,
                                        [width_shift_range, height_shift_range])
    return output_img, label_img

def flip_img(horizontal_flip, vertically_flip, tr_img, label_img):
    if horizontal_flip:
        flip_prob = tf.random.uniform([], 0.0, 1.0)
        tr_img, label_img = tf.cond(tf.less(flip_prob, 0.5),
                                lambda: (tf.image.flip_left_right(tr_img), tf.image.flip_left_right(label_img)), 
                                lambda: (tr_img, label_img))
    if vertically_flip:
        flip_prob = tf.random.uniform([], 0.0, 1.0)
        tr_img, label_img = tf.cond(tf.less(flip_prob, 0.5),
                                lambda: (tf.image.flip_up_down(tr_img), tf.image.flip_up_down(label_img)), 
                                lambda: (tr_img, label_img))
    return tr_img, label_img

def _augment(img,
             label_img,
             resize=None,    # Resize the image to some size e.g. [256, 256]
             scale=1,    # Scale image e.g. 1 / 255.
             hue_delta=0.01,    # Adjust the hue of an RGB image by random factor
             horizontal_flip=True,    # Random left right flip,
             vertically_flip=True,    # Random up down flip,
             width_shift_range=.1,    # Randomly translate the image horizontally
             height_shift_range=.1):    # Randomly translate the image vertically
    
    if resize is not None:
        # Resize both images
        label_img = tf.image.resize(label_img, resize)
        img = tf.image.resize(img, resize)
    
    if hue_delta:
        img = tf.image.random_hue(img, hue_delta)
    
    img, label_img = flip_img(horizontal_flip, vertically_flip, img, label_img)                               
    img, label_img = shift_img(img, label_img, width_shift_range, height_shift_range)
    label_img = tf.cast(label_img, dtype=tf.float32) * scale
    img = tf.cast(img, dtype=tf.float32) * scale
    return img, label_img

data_filenames = list(all_dic.keys())
data_filenames.sort()

def _process_pathnames(filename):
    # We map this function onto each pathname pair
    fname, labels = getData(filename)
    
    img_str = tf.io.read_file(fname)
    img = tf.image.decode_png(img_str, channels=3) # RGB
    
    label_img = tf.convert_to_tensor(labels)
    width = tf.cast(label_img.shape[0], dtype=tf.uint16)
    height = tf.cast(label_img.shape[1], dtype=tf.uint16)
    
    resize = [288, 512]
    img = tf.image.resize(img, resize)
    label_img = tf.image.resize(label_img, resize)
    
    label_cls = tf.reduce_sum(label_img,[0,1])>0
    label_cls = tf.cast(label_cls, dtype=tf.float32)
    
    scale = 1 / 255.                
    img = tf.cast(img, dtype=tf.float32) * scale
    filename = tf.cast(filename, dtype=tf.string)
    
    return filename, width, height, img, label_img, label_cls

#dataset = list(map(_process_pathnames, data_filenames[i]))
#X_train = np.array([x for (x, y) in dataset])
#y_train = np.array([y for (x, y) in dataset])
#print(X_train.shape)

val_ratio = 0.2
seed = 1234
    
train_filenames, test_filenames = train_test_split(data_filenames,
                                                   test_size=val_ratio,
                                                   random_state=seed,
                                                  )

num_train_examples = len(train_filenames)
num_test_examples = len(test_filenames)

import functools
    
img_shape = (288, 512, 3)
    
batch_size = 16
max_epochs = 20
    
def gen_train():
    for i in train_filenames:
        yield _process_pathnames(i)
    
def gen_test():
    for i in test_filenames:
        yield _process_pathnames(i)
    
def get_dataset(#data,
                 preproc_fn = functools.partial(_augment),
                 threads = 6,
                 batch_size = batch_size,
                 is_train = True):
    
    if is_train:
        dataset = tf.data.Dataset.from_generator(gen_train, (tf.string, tf.uint16, tf.uint16, tf.float32, tf.float32, tf.float32), ((), (), (), (288,512,3), (288,512,9), (9)))
    else:
        dataset = tf.data.Dataset.from_generator(gen_test, (tf.string, tf.uint16, tf.uint16, tf.float32, tf.float32, tf.float32), ((), (), (), (288,512,3), (288,512,9), (9)))
    
    if is_train:
        dataset = dataset.shuffle(buffer_size=400)
    dataset = dataset.batch(batch_size)

#    train_dataset, test_dataset = train_test_split(dataset,
#                                                       test_size=val_ratio,
#                                                       random_state=seed,
#                                                      )
    # Map our preprocessing function to every element in our dataset, taking
    # advantage of multithreading
    #dataset = dataset.map(_process_pathnames, num_parallel_calls=threads)
    
#    if is_train:
        #if preproc_fn.keywords is not None and 'resize' not in preproc_fn.keywords:
        #  assert batch_size == 1, "Batching images must be of the same size"
#        dataset = dataset.map(preproc_fn, num_parallel_calls=threads)
#        dataset = dataset.shuffle(num_x * 10)
        
    return dataset

train_dataset = get_dataset(is_train=True)
test_dataset = get_dataset(is_train=False)

print(train_dataset)

<BatchDataset shapes: ((None,), (None,), (None,), (None, 288, 512, 3), (None, 288, 512, 9), (None, 9)), types: (tf.string, tf.uint16, tf.uint16, tf.float32, tf.float32, tf.float32)>

'''
for images, labels in train_dataset.take(1):
    # Running next element in our graph will produce a batch of images
    print(images.shape)
    img = images[0]
    
    plt.figure(figsize=(4,3))
    plt.subplot(1, 1, 1)
    plt.imshow(img)
    plt.show()
    
    plt.figure(figsize=(8,6))
    for i in range(9):
        plt.subplot(3, 3, i+1)
        plt.imshow(labels[0, :, :, i])
    plt.show()
'''

'\nfor images, labels in train_dataset.take(1):\n    # Running next element in our graph will produce a batch of images\n    print(images.shape)\n    img = images[0]\n    \n    plt.figure(figsize=(4,3))\n    plt.subplot(1, 1, 1)\n    plt.imshow(img)\n    plt.show()\n    \n    plt.figure(figsize=(8,6))\n    for i in range(9):\n        plt.subplot(3, 3, i+1)\n        plt.imshow(labels[0, :, :, i])\n    plt.show()\n'

from tensorflow.keras import layers
from tensorflow.keras import models
from tensorflow.keras import losses

class Conv(tf.keras.layers.Layer):
    def __init__(self, num_filters, kernel_size):
        super(Conv, self).__init__()
        self.conv = layers.Conv2D(num_filters, kernel_size, padding='same')
        self.bn = layers.BatchNormalization()

    def call(self, inputs):
        x = self.conv(inputs)
        x = self.bn(x)
        x = layers.ReLU()(x)

        return x

class ConvBlock(tf.keras.layers.Layer):
    def __init__(self, num_filters):
        super(ConvBlock, self).__init__()
        self.conv1 = Conv(num_filters, 3)
        self.conv2 = Conv(num_filters, 3)

    def call(self, inputs):
        conv_block = self.conv1(inputs)
        conv_block = self.conv2(conv_block)

        return conv_block

class EncoderBlock(tf.keras.layers.Layer):
    def __init__(self, num_filters):
        super(EncoderBlock, self).__init__()
        self.conv_block = ConvBlock(num_filters)
        self.encoder_pool = layers.Conv2D(num_filters, 3, padding='same', strides=2, activation='relu')
        # layers.MaxPool2D((2,2))

    def call(self, inputs):
        encoder = self.conv_block(inputs)
        encoder_pool = self.encoder_pool(encoder)

        return encoder_pool, encoder                                         


class DecoderBlock(tf.keras.layers.Layer):                                         
    def __init__(self, num_filters):
        super(DecoderBlock, self).__init__()
        self.convT = layers.Conv2DTranspose(num_filters, 3, strides=2, padding='same')
        self.bn = layers.BatchNormalization()
        self.conv_block = ConvBlock(num_filters)

    def call(self, input_tensor, concat_tensor):
        decoder = self.convT(input_tensor)
        decoder = self.bn(decoder)
        decoder = layers.ReLU()(decoder)
        decoder = layers.concatenate([decoder,concat_tensor]) # tf.concat([decoder, concat_tensor], -1)
        decoder = self.conv_block(decoder)

        return decoder

class UNet(tf.keras.Model):
    def __init__(self):
        super(UNet, self).__init__()
        self.encoder_block1 = EncoderBlock(32)
        self.encoder_block2 = EncoderBlock(64)
        self.encoder_block3 = EncoderBlock(128)
        self.encoder_block4 = EncoderBlock(256)
        
        self.center = ConvBlock(512)
        
        self.flatten = tf.keras.layers.Flatten()
        self.output_linear = tf.keras.layers.Dense(9, activation='sigmoid')
        
        self.decoder_block4 = DecoderBlock(256)
        self.decoder_block3 = DecoderBlock(128)
        self.decoder_block2 = DecoderBlock(64) 
        self.decoder_block1 = DecoderBlock(32) 

        self.output_conv = layers.Conv2D(9, 1, padding='same', activation='sigmoid')

    def call(self, inputs):
        encoder1_pool, encoder1_skip = self.encoder_block1(inputs)
        encoder2_pool, encoder2_skip = self.encoder_block2(encoder1_pool)
        encoder3_pool, encoder3_skip = self.encoder_block3(encoder2_pool)
        encoder4_pool, encoder4_skip = self.encoder_block4(encoder3_pool)
        
        center = self.center(encoder4_pool)
        cls = self.output_linear(self.flatten(center))
        
        decoder4 = self.decoder_block4(center,encoder4_skip)
        decoder3 = self.decoder_block3(decoder4,encoder3_skip)
        decoder2 = self.decoder_block2(decoder3,encoder2_skip)
        decoder1 = self.decoder_block1(decoder2,encoder1_skip)

        outputs = self.output_conv(decoder1)
        
        
        return outputs, cls

model = UNet()

outputs, cls = model(tf.random.normal([batch_size,288,512,3]))
print(outputs.shape, cls.shape)
model.summary()

(16, 288, 512, 9) (16, 9)
Model: "u_net"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
encoder_block (EncoderBlock) multiple                  19648     
_________________________________________________________________
encoder_block_1 (EncoderBloc multiple                  92864     
_________________________________________________________________
encoder_block_2 (EncoderBloc multiple                  370048    
_________________________________________________________________
encoder_block_3 (EncoderBloc multiple                  1477376   
_________________________________________________________________
conv_block_4 (ConvBlock)     multiple                  3544064   
_________________________________________________________________
flatten (Flatten)            multiple                  0         
_________________________________________________________________
dense (Dense)                multiple                  2654217   
_________________________________________________________________
decoder_block (DecoderBlock) multiple                  2952960   
_________________________________________________________________
decoder_block_1 (DecoderBloc multiple                  739200    
_________________________________________________________________
decoder_block_2 (DecoderBloc multiple                  185280    
_________________________________________________________________
decoder_block_3 (DecoderBloc multiple                  46560     
_________________________________________________________________
conv2d_22 (Conv2D)           multiple                  297       
=================================================================
Total params: 12,082,514
Trainable params: 12,075,666
Non-trainable params: 6,848
_________________________________________________________________

def dice_coeff(y_true, y_pred): 
    smooth = 1e-10        
    # Flatten
    y_true_f = tf.reshape(y_true, [-1])
    y_pred_f = tf.reshape(y_pred, [-1])
    intersection = tf.reduce_sum(y_true_f * y_pred_f)
    score = (2. * intersection + smooth) / (tf.reduce_sum(tf.square(y_true_f)) + \
                                            tf.reduce_sum(tf.square(y_pred_f)) + smooth)

    return score

def dice_loss(y_true, y_pred):
    loss = 1 - dice_coeff(y_true, y_pred)
    return loss

def bce_dice_loss(y_true, y_pred):
    loss = tf.reduce_mean(losses.binary_crossentropy(y_true, y_pred)) + \
          dice_loss(y_true, y_pred)

    return loss

bce_loss = tf.keras.losses.BinaryCrossentropy()

from sklearn.metrics import confusion_matrix
from sklearn.metrics import roc_curve
from sklearn.metrics import auc

def metric_function(y_true, y_true_cls, y_pred, y_pred_cls, num_classes=2):
    # Flatten  
    
    y_true_cls = tf.cast(tf.round(y_true_cls), dtype=tf.int32)
    y_pred_cls = tf.cast(tf.round(y_pred_cls), dtype=tf.int32)

    y_true_cls = y_true_cls.numpy().reshape(-1)
    y_pred_cls = y_pred_cls.numpy().reshape(-1)

    y_true_f = tf.reshape(y_true, [-1])
    y_pred_f = tf.reshape(y_pred, [-1])

    y_true_f = tf.cast(tf.round(y_true_f), dtype=tf.int32).numpy()
    y_pred_f = tf.cast(tf.round(y_pred_f), dtype=tf.int32).numpy()
    
    # calculate confusion matrix
    labels = list(range(num_classes))
    current = confusion_matrix(y_true_f, y_pred_f, labels=labels)
    
    # compute mean iou
    intersection = np.diag(current)
    ground_truth_set = current.sum(axis=1)
    predicted_set = current.sum(axis=0)
    union = ground_truth_set + predicted_set - intersection
        
    IoU = intersection / union.astype(np.float32)
    
    sensitivity = []
    specificity = []
    auc_metric = []
    
    # calculate roc curve
    fpr, tpr, thresholds = roc_curve(y_true_cls, y_pred_cls)  
    auc_metric.append(auc(fpr, tpr))

    # calculate confusion matrix
    tn, fp, fn, tp  = confusion_matrix(y_true_cls, y_pred_cls, labels=labels).ravel()
    sensitivity.append(tp/(tp+fn+1e-6))#(tpr)
    specificity.append(tn/(tn+fp+1e-6))#(1-fpr)
    
    sensitivity = np.mean(np.array(sensitivity))
    specificity = np.mean(np.array(specificity))
    auc_metric = np.mean(np.array(auc_metric))
        
    return sensitivity, specificity, auc_metric, np.mean(IoU)

#tf.train.latest_checkpoint(checkpoint_prefix)

optimizer = tf.keras.optimizers.Adam(1e-4)

checkpoint = tf.train.Checkpoint(optimizer=optimizer,
                               model=model)

checkpoint_prefix = '/tf/backup/ckpt/U-Net/'

checkpoint.restore('/tf/backup/ckpt/U-Net/ckpt-3')

<tensorflow.python.training.tracking.util.CheckpointLoadStatus at 0x7fa8d02ca5b0>

#학습코드
'''
from IPython.display import clear_output
    
# save loss values for plot
loss_history = []
metric_history = []
global_step = 0 # step 수 정의 (선택)
print_count = 10 # tf.gradient_tape
save_epochs = 5 # tf.gradient_tape
batch_step = (num_train_examples // batch_size) + (0 if num_train_examples % batch_size == 0 else 1)
    
for epoch in range(max_epochs):
    for filename, width, height, images, labels, cls in train_dataset: # 데이터 로드 파트
        start_time = time.time()
        global_step = global_step + 1
        
        with tf.GradientTape() as tape: # 모델 학습 파트
            predictions, pred_cls = model(images, training=True) # [batch_size, 256,256,3]
            # [batch_size, 256, 256, 1] - prediction
            loss = bce_dice_loss(labels, predictions) # label [batch_size, 256, 256, 1]
            loss = bce_dice_loss(labels, predictions * tf.reshape(tf.round(pred_cls),[pred_cls.shape[0],1,1,9]))
            loss += bce_loss(cls, pred_cls)
        # 가중치 업데이트 파트
        gradients = tape.gradient(loss, model.trainable_variables)
        optimizer.apply_gradients(zip(gradients, model.trainable_variables))
        
        # 학습 상태 출력
        epochs = global_step * batch_size / float(num_train_examples)
        duration = time.time() - start_time

        if ((global_step % batch_step) % (batch_step//print_count)) == 0:
            clear_output(wait=True)
    #        examples_per_sec = batch_size  / float(duration)
            print("Epochs: {:.2f} global_step: {} loss: {:.3f} ({:.3f} sec/batch, {:.3f} sec/epoch)".format(
                        epochs, global_step, loss, duration, duration * num_train_examples / float(batch_size)))
            
            loss_history.append([epochs, loss])

            mean = []

            # print sample image                     
            for filename, width, height, test_images, test_labels, test_cls in test_dataset.take(1):
                sample = 0
                print(filename.numpy()[sample].decode('utf-8').rpartition('.')[0]+'.png')
                predictions, pred_cls = model(test_images, training=False)
                
                print('cls :', test_cls.numpy()[sample])
                print('pred :', pred_cls.numpy().round()[sample])
                
                plt.figure(figsize=(4, 3))
                plt.subplot(1, 1, 1)
                plt.imshow(test_images[sample,:,:,:])
                plt.title("Input image")
                plt.show()

                plt.figure(figsize=(8, 6))
                for i in range(9):
                    plt.subplot(3, 3, i+1)
                    plt.imshow(test_labels[sample,:,:,i])
                    plt.title("Actual Mask "+str(i+1))
                plt.show()

                plt.figure(figsize=(8, 6))
                for i in range(9):
                    plt.subplot(3, 3, i+1)
                    plt.imshow(predictions[sample,:,:,i])
                    plt.title("Predicted Mask "+str(i+1))
                plt.show()
                
                plt.figure(figsize=(8, 6))
                for i in range(9):
                    plt.subplot(3, 3, i+1)
                    plt.imshow(tf.round(predictions[sample,:,:,i]) * tf.round(pred_cls[sample,i]))
                    plt.title("Classification Mask "+str(i+1))
                plt.show()
                _, _, _, c_iou = metric_function(test_labels, test_cls, tf.round(predictions) * tf.reshape(tf.round(pred_cls),[pred_cls.shape[0],1,1,9]), pred_cls)
                
                sen, spe, m_auc, iou = metric_function(test_labels, test_cls, predictions, pred_cls)
                mean.append([sen, spe, m_auc, iou, c_iou])
                
            mean = np.array(mean)
            mean = np.mean(mean,axis=0)
            
            metric_history.append([epochs, mean[0], mean[1], mean[2], mean[3], mean[4]])
            
            score = mean[3]*0.6 + (mean[0]*0.1 + mean[1]*0.1 + mean[2]*0.2)
            real_score = mean[4]*0.6 + (mean[0]*0.1 + mean[1]*0.1 + mean[2]*0.2)
            print("metrics: {}".format(mean))
            print("scores : {}".format(score))
            print("real_score : {}".format(real_score))

    # saving (checkpoint) the model periodically
    if (epoch+1) % save_epochs == 0:
        checkpoint.save(checkpoint_prefix)
'''

#Loss 그래프
'''
loss_history = np.asarray(loss_history)
plt.figure(figsize=(4, 4))
plt.plot(loss_history[:,0], loss_history[:,1])
plt.show()
'''

#Metric 그래프
'''
metric_history = np.asarray(metric_history)
plt.figure(figsize=(12, 8))
plt.subplot(2, 3, 1)
plt.plot(metric_history[:,0], metric_history[:,1])
plt.title('sensitivity')

plt.subplot(2, 3, 2)
plt.plot(metric_history[:,0], metric_history[:,2])
plt.title('specificity')

plt.subplot(2, 3, 3)
plt.plot(metric_history[:,0], metric_history[:,3])
plt.title('AUC')

plt.subplot(2, 3, 4)
plt.plot(metric_history[:,0], metric_history[:,4])
plt.title('iou')

plt.subplot(2, 3, 5)
plt.plot(metric_history[:,0], metric_history[:,5])
plt.title('real_iou')

plt.subplot(2, 3, 6)
plt.plot(metric_history[:,0], metric_history[:,5]*0.6 + metric_history[:,1]*0.1 + metric_history[:,2]*0.1 + metric_history[:,3]*0.2)
plt.title('real_score')
plt.show()
'''

import cv2
import pandas as pd

mean = []

# print sample image                     
for dsf, (filename, width, height, images, labels, cls) in enumerate(test_dataset):
    predictions, pred_cls = model(images, training=False)
    
    mask = np.zeros((predictions.shape[0],288,512,9))
    
    for sample in range(predictions.shape[0]):
        
        no_found = True
        for idx, img_cls in enumerate(pred_cls.numpy().round()[sample]):
            sample_label = predictions[sample].numpy().round()[:,:,idx]

            if(img_cls==1 and sample_label.sum()>500):
                #img = tf.image.resize(np.expand_dims(sample_label, axis=-1), [width.numpy()[sample], height.numpy()[sample]])
                #img = img.numpy()
                
                '''
                plt.figure(figsize=(6, 3))
                
                plt.subplot(1, 1, 1)
                plt.imshow(img, cmap='gray')
                plt.title("Input image")
                '''
                img = (sample_label*255).astype(np.uint8)
                
                contours, hierarchy = cv2.findContours(img, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
                
                maximum = 0
                maximum_iter = 0
                for i, cnt in enumerate(contours):
                    if(maximum < cv2.contourArea(cnt)):
                        maximum = cv2.contourArea(cnt)
                        maximum_iter = i
                points = contours[maximum_iter]

                for i in range(100):
                    pl = cv2.approxPolyDP(points, (i/100)*cv2.arcLength(points, True), True)
                    if(pl.shape[0]<=10):
                        break
                pl = pl.squeeze()

                for i in range(10-len(pl)):
                    pl = np.concatenate((pl, pl[0].reshape(1,2)), axis=0)
                '''
                plt.plot(pl[:,0],pl[:,1])
                plt.show()
                '''
                pl = [(x[0],x[1]) for x in pl]
                _img = Image.new('L', (512,288), 0)
                ImageDraw.Draw(_img).polygon(pl, outline=1, fill=1)
                mask[sample,:,:,idx] = np.array(_img)#tf.image.resize(np.expand_dims(np.array(_img), axis=-1),[288,512]).numpy().squeeze()
    '''
    print('---------------------')
    print('pred  :', tf.round(pred_cls).numpy()[-1])
    print('class :', cls.numpy()[-1])
    plt.figure(figsize=(18, 4))
    for i in range(9):
        plt.subplot(2, 9, i+1)
        plt.imshow(mask[-1,:,:,i],cmap='gray')
        plt.title("Predicted Mask "+str(i+1))
        
    for i in range(9):
        plt.subplot(2, 9, 9+i+1)
        plt.imshow(labels[-1,:,:,i],cmap='gray')
        plt.title("Actual Mask "+str(i+1))
    plt.show()
    '''
    sen, spe, m_auc, iou = metric_function(labels, cls, mask, tf.round(pred_cls).numpy())
    mean.append(sen*0.1 + spe*0.1 + m_auc*0.2 + iou*0.6)

#   print('sen :',sen, 'spe :', spe, 'auc :', m_auc, 'iou :', iou)
    print('score :',sen*0.1 + spe*0.1 + m_auc*0.2 + iou*0.6)

score : 0.8643053661958066
score : 0.845951569638252
score : 0.7972379591691043
score : 0.8541119383124611
score : 0.8164731811676046
score : 0.7862118250548519
score : 0.8904176039133345
score : 0.9252477283026014
score : 0.8860444935320874
score : 0.8707181043936458
score : 0.8808662974249755
score : 0.8544056816536663
score : 0.8407920177197137
score : 0.850172060369548
score : 0.8735735810389491
score : 0.8138009199083058
score : 0.8671103443686123

---------------------------------------------------------------------------
KeyboardInterrupt                         Traceback (most recent call last)
<ipython-input-53-4f3d234d928b> in <module>
     71     plt.show()
     72     '''
---> 73     sen, spe, m_auc, iou = metric_function(labels, cls, mask, tf.round(pred_cls).numpy())
     74     mean.append(sen*0.1 + spe*0.1 + m_auc*0.2 + iou*0.6)
     75 

<ipython-input-49-f3fa0ddc1c4f> in metric_function(y_true, y_true_cls, y_pred, y_pred_cls, num_classes)
     16 
     17     y_true_f = tf.cast(tf.round(y_true_f), dtype=tf.int32).numpy()
---> 18     y_pred_f = tf.cast(tf.round(y_pred_f), dtype=tf.int32).numpy()
     19 
     20     # calculate confusion matrix

/usr/local/lib/python3.8/dist-packages/tensorflow/python/framework/ops.py in numpy(self)
   1092     """
   1093     # TODO(slebedev): Consider avoiding a copy for non-CPU or remote tensors.
-> 1094     maybe_arr = self._numpy()  # pylint: disable=protected-access
   1095     return maybe_arr.copy() if isinstance(maybe_arr, np.ndarray) else maybe_arr
   1096 

/usr/local/lib/python3.8/dist-packages/tensorflow/python/framework/ops.py in _numpy(self)
   1058   def _numpy(self):
   1059     try:
-> 1060       return self._numpy_internal()
   1061     except core._NotOkStatusException as e:  # pylint: disable=protected-access
   1062       six.raise_from(core._status_to_exception(e.code, e.message), None)  # pylint: disable=protected-access

KeyboardInterrupt: 

#테스트 스코어
np.mean(np.array(mean))

inference_list = os.listdir(inference_data)
inference_list.sort()
inference_list

def _process_inference(filename):
    # We map this function onto each pathname pair
    img_path = inference_data+'/'+filename
    
    img_str = tf.io.read_file(img_path)
    img = tf.image.decode_png(img_str, channels=3) # RGB
    
    width = tf.cast(img.shape[0], dtype=tf.uint16)
    height = tf.cast(img.shape[1], dtype=tf.uint16)
    
    resize = [288, 512]
    img = tf.image.resize(img, resize)
    
    scale = 1 / 255.                
    img = tf.cast(img, dtype=tf.float32) * scale
    filename = tf.cast(filename, dtype=tf.string)
    
    return filename, width, height, img

def gen_infer():
    for i in inference_list:
        yield _process_inference(i)
    
def dataset_infer(batch_size=16):
    
    dataset = tf.data.Dataset.from_generator(gen_infer, (tf.string, tf.uint16, tf.uint16, tf.float32), ((), (), (), (288,512,3)))
    
    dataset = dataset.batch(batch_size)
    return dataset

infer_dataset = dataset_infer()

import cv2
import pandas as pd

results = pd.DataFrame(columns=['file', 'class', 'x1', 'y1', 'x2', 'y2', 'x3', 'y3', 'x4', 'y4', 'x5', 'y5', 'x6', 'y6', 'x7', 'y7', 'x8', 'y8', 'x9', 'y9', 'x10', 'y10'])
count = 0

# print sample image                     
for filename, width, height, images in infer_dataset:
    predictions, pred_cls = model(images, training=False)
    for sample in range(predictions.shape[0]):
        no_found = True
        for idx, img_cls in enumerate(pred_cls.numpy().round()[sample]):
            sample_label = predictions[sample].numpy().round()[:,:,idx]
            if(img_cls==1 and sample_label.sum()>300):
                img = tf.image.resize(np.expand_dims(sample_label, axis=-1), [width.numpy()[sample], height.numpy()[sample]])
                img = img.numpy()
                '''
                plt.figure(figsize=(6, 3))
                
                plt.subplot(1, 1, 1)
                plt.imshow(img, cmap='gray')
                plt.title("Input image")
                '''
                img = (img*255).astype(np.uint8)
                
                contours, hierarchy = cv2.findContours(img, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
                
                maximum = 0
                maximum_iter = 0
                for i, cnt in enumerate(contours):
                    if(maximum < cv2.contourArea(cnt)):
                        maximum = cv2.contourArea(cnt)
                        maximum_iter = i
                points = contours[maximum_iter]

                for i in range(100):
                    pl = cv2.approxPolyDP(points, (i/100)*cv2.arcLength(points, True), True)
                    if(pl.shape[0]<=10):
                        break
                pl = pl.squeeze()

                for i in range(10-len(pl)):
                    pl = np.concatenate((pl, pl[0].reshape(1,2)), axis=0)
                '''
                plt.plot(pl[:,0],pl[:,1])
                plt.show()
                '''
                pl = pl.flatten()
                result = [filename.numpy()[sample].decode('utf-8').rpartition('.')[0], idx]
                result.extend(pl)
                #print(result)
                results.loc[count] = result
                no_found = False
                count += 1
        if(no_found):
            result = [filename.numpy()[sample].decode('utf-8').rpartition('.')[0]]
            result.extend([None] * 21)
            results.loc[count] = result
            count += 1

results

results.to_csv('results.csv',index = False)

# threshold 정하기 위해 알아본 함수
'''
a = []
for train_images, train_labels in train_dataset:
    for i in range(len(train_labels)):
        temp = (train_labels[i,:,:,:].numpy().round().sum(axis=(0,1)).max())
        if(temp < 2000):
            print(temp)
            a.append(temp)
print(np.array(a).min())
'''